The SSD genie is out of the bottle. It is common knowledge that Solid State Drives (SSD) are many times faster than traditional hard disk drives (HDD) – which is why enterprise-grade SSD and flash storage arrays are rapidly replacing traditional HDD-based storage solutions as the de facto standard for time sensitive, tier one applications. However, conventional wisdom dictates that SSD or flash is much more expensive. In fact, when performance counts, Flash is not only faster, in many cases, Flash is a lot less expensive than HDD.
A performance-centric data storage total cost of ownership (TCO) model recently developed by Parity Research demonstrates why flash arrays are much more cost-effective than HDD arrays. Parity compared an all-flash array from WHIPTAIL to an all-HDD array. (WHIPTAIL acquired by Cisco in late 2013)
Some highlights of the study include the following TCO assumptions, infographic and key findings.
Contents |
TCO Assumptions
- All necessary additional hardware and software costs bundled into pricing, including servers, switches, storage management software, racks, cabling, enclosures, storage operating systems, and legacy storage interfaces (approximate cost: $30k to $40k).
- Both performance arrays calculated at minimum 250k IOPS (400 IOPS per 15k RPM 150gb HHD and 4,000-plus IOPS for 150gb SSDs).
- 70% utilization rate for flash array and 30% for all-HDD array includes data protection levels up to RAID 6.
- $800 per square foot of data center rack space assumes a highly efficient data center with average availability needs (5 sq. feet per 2U rack).
- 200 watts per fully populated SSD enclosure, 400 watts for HDD primary enclosure and 200 for additional enclosures (cooling requires 1.2x number of watts to power. Cost of power calculation = .12/kwh).
- Maintenance and drive longevity assumed to be the same.
- Admin cost higher for HDD due to more components and additional tuning required.
- Labor costs and training assumed to be the same for both systems.
- Warranties and extended warranties assumed to be the same price.
Key Findings
- HDD array costs over $500,000 more to maintain than SSD array over five years—almost equal to the cost of an additional, similarly configured flash array.
- HDD array takes up more than eight-times the physical space of a flash array (24 disks per enclosure for both HDD and SSD).
- HDD array consumes almost 10 times the power of flash array.
- Administrative costs are almost 500% higher with HDD array due to more disks (625 HDDs vs. 72 SSD) and added complexity, including additional management software and tuning, more wiring, mechanical components, and more points of failure.
- HDD deployment costs are more than five-times the cost of deploying SSD flash.
- Flash array has more IOPS headroom for consolidating performance-centric application workloads, assuring required service levels.
HDD TCO Considerations
- 60% or more of HDDs spin empty all the time.
- Companies may have to spend upwards of $1 million to manage tiered or hybrid storage arrays.
- Enterprise HDD storage can account for up to 6x the power consumption of servers
- Since its introduction by IBM in 1956, basic HDD mechanics have not changed
Conclusion
While the TCO scenario depicted in the infographic is subject to a number of factors that would swing the cost advantage one way or the other, for today’s increasingly IOPS-intensive workloads, flash arrays are definitively more cost-effective and provide much higher performance than HDD performance arrays. HDD arrays will continue to provide value for applications where multiple petabyte scale capacity is needed or where speed and response time is not an issue, such as with archiving documents and other data that is infrequently accessed.
When application performance is the primary goal, flash arrays are a far superior choice to HDD arrays on every level, including price.
Action Item: CIOs, application owners and data storage administrators need to review their primary storage acquisition strategies and, if they haven't already, include SSD and flash arrays in their storage portfolio to decrease latency and improve responsiveness of time-critical applications while simultaneously decreasing the TCO for their performance-centric data storage assets.
Footnotes: TCO research and methodology developed by Gary MacFadden of Parity Research with assistance from Cal Braunstein, CEO of The Robert Frances Group
(Whiptail acquired by Cisco in late 2013)